Prognostic Value of Coronary Computed Tomography Angiography in Patients With Diabetes: A Meta-analysis

Diabetes is associated with an increased risk of coronary atherosclerosis and excess cardiovascular morbidity and mortality (1). Atherosclerosis in patients with diabetes manifests in a more accelerated and progressive manner. Overall, a twofold risk for developing coronary artery disease (CAD) has been observed in this patient population (2). Patients with diabetes may have a similar risk for new-onset myocardial infarction as patients without diabetes with prior myocardial infarction (3). Cardiac stress testing is considered appropriate for the identification of CAD in symptomatic patients with an intermediate or high risk (4). However, the role of stress imaging in asymptomatic patients remains controversial (5,6).

In recent years, coronary computed tomography angiography (CTA) has emerged as a reliable noninvasive imaging tool for the identification of CAD. CTA allows for the precise evaluation of the coronary lumen and, with a high negative predictive value, enables CAD to be ruled out (7,8). Several studies have investigated the prognostic value of CTA for the prediction of future cardiac events in patients with diabetes. Results suggest that CTA provides incremental prognostic information for suspected CAD and that the absence of CAD is associated with an excellent prognosis (9,10). However, a recent randomized clinical trial, which was not adequately powered, showed that CTA screening for CAD in patients with diabetes did not reduce the composite rate of all-cause mortality, nonfatal myocardial infarction, or unstable angina requiring hospitalization during a 4-year follow up (11). Recent guidelines urge caution about using CTA as a first-line imaging tool for the identification of CAD, stating CTA “may be” rather than “is” appropriate in patients with diabetes (4). The evidence from individual CTA studies is limited due to differences in pretest probability and event definition. Hence, the actual risk associated with obstructive and nonobstructive CAD and the usefulness of CTA in this patient population remains ambiguous. We therefore performed a meta-analysis of studies reporting event rates and hazard ratios (HR) to determine the prognostic value of CTA in this patient population.

This meta-analysis was conducted using the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) Guidelines (12). The study protocol was published online (www.crd.york.ac.uk/PROSPERO; PROSPERO 2015: CRD42015029218). A predefined search syntax was used to find eligible studies in PubMed and Embase up to 12 November 2015 (Supplementary Table 1). No search or language restrictions were imposed. Unpublished studies were not included, and no attempt was made to contact authors. A manual reference check of full-text articles was performed to identify studies missed by our systematic search.

Two physician–scientists (C.C. and R.A.P.T.) independently selected articles from both databases using predefined inclusion and exclusion criteria (Fig. 1A). Eligibility criteria were study domain, patients with diabetes suspected of CAD; index test, coronary CTA, obstructive CAD (defined as ≥50% luminal narrowing), nonobstructive CAD; events (all-cause mortality or cardiac death, nonfatal myocardial infarction, unstable angina pectoris, stroke, revascularization) or events excluding revascularization and unadjusted and/or adjusted HR for events and events excluding revascularization. If there was an overlap in study populations, the study with the largest population was included. Studies with fewer than 10 events and abstracts were not incorporated in the meta-analysis.

Characteristics and outcomes (including unadjusted/adjusted HR) of study subjects were collected by one physician–scientist and checked by another for all included studies. Prevalences and weighted annualized event rates (number of events divided by median follow-up time and weighted by sample size) were calculated. Study quality was ascertained in consensus using the Quality In Prognosis Studies (QUIPS) tool (13).

All statistical analyses were performed on a patient level. Naturally log-transformed HR and standard errors were calculated to pool the HR. A random-effects generic inverse variance method was used to generate pooled HR. A funnel plot was generated to graphically determine publication bias. Forest plots were created to graphically display HR. Heterogeneity was quantified calculating the I² statistic. The degree of heterogeneity was considered low (I² < 50%), moderate (I² = 50–75%), or high (I² > 75%) (14). Meta-regressions were performed for obstructive CAD to evaluate the effect of revascularization on the HR and to calculate the difference between unadjusted and adjusted HR. Statistical analyses were performed using RevMan 5.3 software (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark) and Stata 12 software (StataCorp LP, College Station, TX).

This meta-analysis included eight studies that met the predefined inclusion criteria (Fig. 1A). Study characteristics are provided in Table 1. The study population consisted of 6,225 participants (56% male; weighted age, 61 years) with diabetes who were referred for coronary CTA for suspected CAD. The overall quality of the prognostic studies was good (Fig. 1B and Supplementary Table 2). One study included participants with known history of CAD (10). Definitions for diabetes were comparable across the studies, and studies were performed in North America, Europe and Asia (Supplementary Table 3). A prospective study design was used for 87.5% of the included studies. The follow-up period of individual studies ranged from 20 to 66 months. In seven studies the evaluation of CAD was performed using ≥64-slice CT scanner, whereas one study used a 16-slice and a 64-slice CT scanner (15).

Eight studies (five with HR adjusted for clinical risk factors, three without adjustment), including 6,225 participants and 616 events (Table 2 and Supplementary Table 4), reported HR for obstructive CAD (Supplementary Table 5). Among the 616 events, 108 (17.5%) were from a retrospective study (16) reporting only all-cause mortality. The seven other studies all had composite events: four reported all-cause mortality, three reported cardiac mortality, five reported unstable angina pectoris, six reported nonfatal myocardial infarction, one reported stroke, and four reported revascularization (Supplementary Table 4). The weighted prevalence of obstructive CAD was 38%, and the weighted annualized event rate was 17.1% (Table 2). Variability in annualized event rate among studies was considerable, ranging from 2.3 to 27.8%. The pooled HR for obstructive CAD was 5.4 (95% CI 3.2–9.0) (Fig. 2A). Heterogeneity was substantial (I² = 69%) among studies. The funnel plot did not reveal sign of asymmetry on visual inspection (Supplementary Fig. 1). When a meta-regression was performed on including/excluding revascularization events, a higher HR was observed for obstructive CAD in studies that included revascularization compared with those that did not (7.3 [95% CI 5.1–10.4] vs. 3.7 [95% CI 1.8–7.4], P = 0.124). When unadjusted and adjusted HRs were compared for obstructive CAD, the unadjusted HR was higher than the adjusted HR (7.4 [95% CI 5.0–11.1] vs. 4.7 [95% CI 2.0–10.7], P = 0.319).

Three studies (one unadjusted), consisting of 4,196 patients and 195 events, reported HR for nonobstructive CAD (Supplementary Table 5). The weighted prevalence of nonobstructive CAD was 36%, and the weighted annualized event rate was 4.5% (Table 2). Variability in the annualized event rate was substantial, ranging from 1.0 to 9.3%. The pooled HR for nonobstructive CAD was 4.2 (95% CI 2.3–7.6) (Fig. 2B). Heterogeneity was low (I ² = 0%). Owing to the low number of studies reporting HR for nonobstructive CAD, no further subanalyses were performed. In patients with absence of CAD, the weighted prevalence was 25% and the weighted annualized event rate was 0.1%.

To our knowledge, this is the first meta-analysis that reports the prognostic value of coronary CTA in patients with diabetes. Our results indicate that in patients with diabetes, the presence of obstructive and nonobstructive CAD on CTA is associated with an increased risk of mortality and cardiovascular events. Specifically, obstructive CAD on CTA was associated with a fivefold and nonobstructive CAD with a fourfold risk for events. It is well known that diabetes is associated with an increased risk of cardiovascular morbidity and mortality (1,17) and that diabetes has been regarded to be risk equivalent to CAD (18,19). Detecting CAD in patients with diabetes is challenging (20). The involvement of small vessels due to metabolic abnormalities and the diffuse nature of the disease limit the reliability of cardiac stress tests for detecting myocardial ischemia (21). In addition, the silent fashion of CAD due to the high threshold for pain reduces the sensitivity of clinical risk assessment (22). The American Diabetes Association and American Heart Association recently issued a joint statement that urges the identification of asymptomatic patients with subclinical CAD in whom more aggressive lifestyle or treatment changes would allow prevention of progression of the disease and reduce future clinical events (23). Coronary CTA has become a powerful diagnostic tool for ruling out obstructive CAD (24). It has been also demonstrated that the presence of CAD on CTA in patients with diabetes is associated with worse outcome, whereas the absence of CAD shows an excellent prognosis (9,10). The prevalence of obstructive CAD in our meta-analysis was 38%, and the presence of obstructive CAD yielded a weighted annualized event rate of 17.1%. Comparatively, a meta-analysis by Bamberg et al. (25) found in patients with suspected/known CAD (15% with diabetes) a prevalence of obstructive CAD of 29% with a weighted annualized event rate of 11.9%. Hence, our results indicate a greater disease burden in patients with diabetes.

CTA enables the detection of nonobstructive stenosis. We found a high prevalence of nonobstructive CAD in patients with diabetes with a still considerable annualized event rate of 4.5%. Theoretically, nonobstructive CAD cannot be detected by other noninvasive modalities used to identify ischemia (e.g., single-photon emission computed tomography, exercise tolerance test, or stress echocardiography). Moreover, with a prevalence of 25% observed in our meta-analysis, the absence of CAD is a common diagnosis in patients with diabetes, which is associated with a very low event rate (0.1%). This finding demonstrates the role of coronary CTA in safely ruling out future events in patients with diabetes and yielding a similar event rate for absence of CAD as a general patient population referred for CTA (25). Nonetheless, CTA is accompanied by radiation exposure and the administration of iodine-containing contrast material.

Diabetes is associated with a poorer outcome after revascularization (26), especially in-hospital mortality rates in patients undergoing urgent versus elective revascularization (12.7% vs. 1.4%) (27). Screening for CAD in patients with diabetes could enable the identification of high-risk patients in whom event-free survival may be improved through risk factor modification, medical surveillance, or elective revascularization. The FACTOR-64 randomized clinical trial (11) showed that CTA screening of patients with diabetes for CAD did not result in a reduction of all-cause mortality, nonfatal myocardial infarction, or unstable angina requiring hospitalization at 4 years. However, the study was underpowered due to a low event rate (16% anticipated; 7.6% observed in the non-CTA arm) and an optimistically assumed event reduction of 40% within the CTA arm. In addition, care targets for risk factor reduction were not met in most of the patients assigned to aggressive treatment in the CTA group (11), which could have been caused by suboptimal adherence to the prescribed therapy. Long-term optimal management of patients with diabetes remains challenging.

A recent study evaluated data from three comprehensive clinical trials—COURAGE (Clinical Outcomes Utilizing Revascularization and Aggressive Drug Evaluation), BARI 2D (Bypass Angioplasty Revascularization Investigation Type 2 Diabetes) and FREEDOM (Future REvascularization Evaluation in patients with Diabetes mellitus: Optimal management of Multivessel disease)—in an attempt to assess their effectiveness in achieving guideline-driven risk reduction in patients with diabetes and CAD. They found that a low proportion of participants (18%, 23%, and 8%, respectively) did achieve all prespecified treatment targets at 1 year (28), which stresses the need for adequate patient counseling in those assigned to aggressive treatment targets. Results from the BARI 2D trial (29) showed no significant difference in death and major cardiovascular events between optimal medical therapy and early revascularization in patients with diabetes and stable ischemic heart disease, although a significant reduction in nonfatal myocardial infarction was observed in patients allocated to coronary artery bypass grafting.

Our meta-analysis provides insights into the role of coronary CTA in patients with diabetes. Coronary artery calcium scoring (CACS) has been proposed as a first-line test for CAD in patients with diabetes (30). Only two of the included studies (31,32) assessed the prognostic value of stenosis degree on CTA beyond CACS, and both concluded that CTA provided additional value over CACS. In addition, several other studies confirm the incremental value of CTA. Most notably is that a CACS of 0 does not exclude obstructive CAD in a substantial proportion of patients: Park et al. (33) found obstructive CAD in 8.0% and Min et al. (34) in 10.5% of patients with CACS of 0.

An important strength of the present meta-analysis is that we evaluated the prognostic value of CTA in a large number of patients with diabetes, although several limitations deserve mention. First, we did not evaluate coronary CTA beyond the degree of stenosis, and because of heterogeneous reporting, we were also not able to determine the risk associated with triple-vessel disease. Second, treatment bias is likely. CTA leads to the identification of obstructive CAD, likely resulting in increased revascularization rates but also potentially in improved patient survival. Third, heterogeneity was considerable among studies. This is likely caused by differences in pretest probability and in event definition. Finally, the availability of adjusted HR and diabetes features were not uniformly available in all studies.

In conclusion, obstructive and nonobstructive CAD on CTA is associated with increased event rates and increased HR in patients with diabetes. The absence of CAD on CTA is associated with a low event rate. CTA in patients with diabetes allows for safely ruling out future events, and the detection of CAD could allow for the identification of high-risk patients in whom aggressive risk factor modification, medical surveillance, or elective revascularization could potentially improve survival. Large multicenter studies with long follow-up and uniform reporting of HR and diabetic characteristics are still needed to fully comprehend the prognostic value of CTA in diabetes.

Funding. R.A.P.T. is supported by Van Leersum Grant of the Royal Netherlands Academy of Arts and Sciences.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.C. and R.A.P.T. contributed to study conception and design, acquired data, and drafted the article. C.C., P.M.-H., B.B.G., B.M., T.L., and R.A.P.T. contributed to data analysis and interpretation. P.M.-H., B.B.G., B.M., and T.L. contributed to critical revisions of the article for important intellectual content. R.A.P.T. performed statistical analysis.